AI Agents public
Production-grade agent skills and Custom GPT prompts for ChatGPT, Claude Code, and Codex. 140 skills, 28 agents, Agent Skills spec compliant.
Install
/plugin marketplace add vasilyu1983/AI-Agents-publicInstall command taken from the project's README or registry entry. Always review code before running it.
README excerpt
Read full README on GitHub →AI-Agents
Ready-made expertise and multi-agent runs for Claude Code and Codex.
Out of the box, Claude Code works as one assistant in one chat. This repository adds three things:
- Skills: instructions the assistant loads by itself when your task needs them, for example a code-review checklist.
- Subagents: specialists, such as a security reviewer, that work in a separate context and report back.
- Saved workflows: multi-step runs, such as "review my diff with 4 reviewers, then try to disprove each finding". The rules of each run are enforced in code, so an agent cannot skip a step.
Install it as a Claude Code plugin in three commands, or clone it and link it.
| What you get | How many | Where |
|---|---|---|
| Skills: engineering, AI, data, QA, ops, docs, research, applied theory | 140+ | skills/universal/ |
| Subagents with explicit boundaries | 85 | agents/claude/, agents/codex/ |
| Team recipes | 8 | agents/teams/ |
| Saved workflows with guarantees in code | 8 | agents/workflows/ |
| Opt-in hooks: git safety, lint-config guard, notifications | 3 | hooks/ |
| Language and domain rules | 14 folders | rules/ |
The exact, current list is in the generated catalog.
Quick start
Pick one install path. Claude Code only: use the plugin. Codex, or you want to edit skills: clone and link.
Claude Code plugin
/plugin marketplace add vasilyu1983/AI-Agents-public
/plugin install ai-agents@ai-agents
/reload-plugins
Then, in a repository with uncommitted changes:
/ai-agents:adversarial-review
Four reviewers read your diff. Two more agents (refuters) then try to disprove each finding. A finding is dropped only if both disprove it. The run changes no files.
Cost: each reviewer and refuter is a full agent session on your own plan. One run starts 4 reviewers plus 2 refuters per finding, and takes several minutes. Run it on a small diff first.
The run ends with a verdict: CLEAN (no findings), FINDINGS (each with file:line, severity and title), or INCOMPLETE (an agent failed; re-run). Workflows shows the full result.
Clone and link (Claude Code and Codex)
git clone https://github.com/vasilyu1983/AI-Agents-public.git ~/AI-Agents-public
cd ~/AI-Agents-public
bash scripts/distribution/sync-skills.sh
Then start a new session and type /adversarial-review. In Codex, which has no workflow commands, invoke the run-workflow skill and name the workflow; the skill makes Codex follow the workflow step by step.
Getting started covers both paths, subagent installs, checks, updates and removal.
Documentation
Similar plugs
Makes your AI agent think like the laziest senior dev in the room. The best code is the code you never wrote.
Turn any codebase, with its docs, SQL schemas, configs, and PDFs, into a queryable knowledge graph. A /graphify skill for Claude Code, Cursor, Codex, and Gemini CLI: local…
Why use many token when few token do trick. Viral skill + proxy for coding agents that cuts 65% of tokens by talking like a caveman.
Production-grade engineering skills for AI coding agents.
The 1st Collaborative AI Design Agent Workspace. Local-first and open source. Create prototypes, websites, apps, dashboards, presentations, images, and videos with your team…